Coder

Have you used similar configuration and pretrained of Swin-Unet ?

Query

1

hi, can you confirm if the code use the exactly same layers of actual model or there is any variations? Please confirm fast. I am a beginner need your assitance

AssertionError: Invalid type for key QK_SCALE; valid types = {, , , , , }

How to solve it ?

Urgent Query

3

I am confused about the MoE layer in Jamba block. There are many versions of MoE. The paper has not defined in detail the mathematics or diagrams to understand the...

??

Pretrain Query

1

In transunet original code there are not all pretrains defined like for r_50_l_16 or l_32. but you have done it. Will there be any difference comes with this? Please guide

how to solve

In code, there are some unused functions, like conv1x1, AverageMeter, count_param, str2bool, qkv. why they are still kept in the official. I would be grateful for the clarity?